Name: Product Demo | Oscillating Heat Pipe Enhanced Cold Plate | Presented by ThermAvant
Uploaded: 2026-07-22T15:45:52.699Z
Duration: 30 min 52 s
Description: Product Demo | Oscillating Heat Pipe Enhanced Cold Plate | Presented by ThermAvant

Transcript for "Product Demo | Oscillating Heat Pipe Enhanced Cold Plate | Presented by ThermAvant": Welcome back to data center cooling days. Let's keep things moving. In this session, we will learn about a recent innovation from the team at ThermoVont, an oscillating heat pipe enhanced cold plate. Our presenter is doctor Corey Wilson. He's the senior director of research and development at Thermavont, and he brings nearly two decades of experience in two phase thermal fluids and specialized expertise in oscillating heat pipes. Thermovant Technologies designs and manufactures advanced thermal management solutions, specializing in, oscillating heat pipe technology, as you'll hear about, for high power app applications. You can send in your questions for Corey in the q and a tab to the right of the chat area, and we will answer them at the end of Corey's presentation. Now, Corey, I'd like to welcome you to data center cooling days. Hi. Thank you very much. Take it away. Okay. So, today, I'm presenting on the Yasinski pipe enhanced cold plate. A quick summary and then I'll get into the details of this, of this design that we developed. So, I'll explain more about what an Auslan Heat pipe is. But basically, we took an Auslan Heat pipe and integrate it monolithically with a cold plate to spread heat from a small heat high flux heat source, to a, the finned area of the cold plate. So we pushed it up to 1,389 watts, 174 watts per square centimeter, and we rejected heat into well, we tested a few different flow rates, but, you know, this experimental test case here is 2.2 liters per minute with inlet of 35 Celsius. And we are able to achieve a thermal resistance of point zero two calvins per watt or point one six four, calvins and we're squared per watt. So basically, this matches the performance of existing high end cold plates, has substantially lower pressure drop because we are increasing the rejection area into the cold plate and the mass is substantially lower than existing high end cold plates. So Thermobond, we've been around since 2007. We have, grown substantially in the last two decades, and our primary product is the Asli Key pipe. We also developed some single phase pump cold plates and other thermal products. So this oscillating pipe that I was talking about, it's relatively new. It was invented back in 1990, but it was an academic curiosity for a very long time. And it wasn't until about a decade to a decade and a half ago that people were starting to really, develop this commercially, and ThermoDOT has taken the lead on that. Our first on orbit, oscillating heat pipe for an experiment was in 2017, and then our first oscillating heat pipe used on a satellite in a commercial application was in 2021. And since then, we've had over ten million hours on orbit operation. So we primarily service the aerospace industry, but this Auslan heat pipe is broadly useful across a wide range of applications, which is why we're developing it for this cold plate. Okay. What is an Auslan heat pipe? It's a type of heat pipe, but it operates substantially different than other heat pipes if you're familiar with them. The Auslund heat pipe is a, meandering capillary channel. So in this visual, the channel is about a millimeter by a millimeter. And the plate on the left side, you're seeing the interior of the Auslund heat pipe. A lid would be bonded onto the top, so it's an enclosed channel, fully hermetic. And you charge it with a small amount of working fluid, maybe 50% would be, you know, typical amount of working fluid. And that fluid where you heat it, your vapor expands, your, liquid evaporates, and you have nucleation events, some boiling occurring. And where it's cooled, you get some condensation, you get vapor contraction, and this results in a very chaotic, very random motion of the fluid. This nucleation, the nucleation events that are occurring in these pressure gradients push that fluid around transporting heat from the heat source to the heat sink via sensible and latent heat transfers. So it's a very chaotic, basically enclosed boiling in a capillary channel. And in the operating region, which we'll get into in a second, the Aslan heat pipe can transport substantial heat loads in a very thin, form factor. So the Auslin heat pipe, how we typically manufacture there are a few different ways, but we typically manufacture by machining channels into a plate and then, making multiple layers of material and bonding them together into this chromatic monolithic structure. I, did a cross section here so you can see that channel. And then we remove the exterior material to get the final shape for the customer's application, which might include through holes or threaded locations, and then we charge and seal it with our working fluid. Okay. So an Auslan heat pipe, it is charged with a saturated working fluid. So the working fluid for the Auslan heat pipe to work must be between the triple point and the critical point, but there's also other, limitations on its operation. On the right side, I have the conventional heat pipe limits of operation for those who are familiar with them. But the Aslan heat pipe is operates quite a bit differently and therefore has different limits. The startup limit is where nucleation begins in the Aslan heat pipe. Below that, the Austin heat pipe will still transport some heat via convective flow, but the conductance, the overall performance of Austin heat pipe is quite low. Above that start up limit, it starts working really well. So the startup limit is a soft limit. Basically, once you push the heat, the heat load higher or you increase the temperature, the gas and heat pipe starts removing the heat. The sonic limit is, where the fluid moves too fast and therefore, you hit to the speed of sound. Usually, you have these viscous contours where the viscous delta t ends up becoming too substantial for your application, so that's also a pseudo limit. And then at the upper end, we have the vapor inertia limit where the fluid inside the small capillary channel is moving so fast that the flow goes angular. Your liquid is just laid down on the channel wall and the vapor blows through the middle. That, when that occurs, the fluid motion will cease and the oscillating heat pipe will quit working. The swept link limit is also a limit where the oscillating heat pipe will quit working if it exceeds that limit, and that's where your nucleation rate exceeds the flow rate of the fluid and you can't push the fluid through the heater as quick enough. So above those limits, the pipe will quit working and you can have a runway event where your temperatures rise. Below that, the oscillating heat pipe works very well. So, our design that we have developed, which I'll get into in a second, is well within this operating region, well away from those limits. Okay. So this design that we developed was an OHP combined with a cold plate. And how we did that is we built it monolithically, but, the Ossenheat pipe was made out of aluminum sixty sixty one. And the Ossenheat pipe is that green path that you're seeing in those two images, the two cross sections. So the heat comes in, It goes through a thin layer of aluminum, and then it enters the oscillating pipe where this random oscillating motion will rapidly move heat around the interior structure of that green region and then can reject heat into the cold plate region, substantially increasing the rejection area. So in the application that we designed for, we had eight square centimeters of heat input, but the finned area in our cold plate was 41.3 square centimeters. So by using us and heat pipe, we've increased that rejection area into the working fluid 5.2 x from that key input area. So it's a very substantial, increase in the base area of your fin array. Now the design we developed, we're primarily an awesome heat pipe manufacturer. Our fin pattern is relatively simplistic. It has some benefits, which we'll get into in a minute, but our fins were half a millimeter thick with one millimeter gaps between them. So it's relatively coarse, grid pattern. And future designs will focus on fin optimization for various applications. Okay. So we tested this Ausling heat pipe in this test rig shown here. We had attached it to a, a pump system where we could control the flow rates, the inlet temperature, and we used, ethylene glycol water at a sixty forty ratio. I know that for the data center industry, you're typically using PG 25, propylene glycol at 25%, mixture. We used EGW because that's what we typically use in our industry. For the aerospace industry where we need a wide operating temperature range. The this ETW because it's optimized for wide operating temperature range has a higher viscosity than the p two twenty five, a lower thermal conductivity, and a lower specific heat, but it does have a slightly higher density so that, sort of slightly balances with the specific heat. So, basically, the EGW is a lower performance working fluid and will remove less heat. This awesome heat pipe can work with a corrosion inhibited p g 25. We just did not test it that way. The OHP was heated with a cartridge heater inserted into a copper block, concert any heat onto the contact area. And we're able to reach 1,400 watts or 178 watts per square centimeter with this setup, and it could reach a maximum temperature of 80 degrees Celsius. We also are able to test it on a horizontal and a side orientation to prove gravity independence. So the thermocouples, we put many temperature sensors throughout the oscillating heat pipe, but the primary or the most important temperature sensor was directly on the heat source. We had thermocouple on a little, RTV silicone pad to press it onto the, case of that cold plate to get an accurate case temperature. And other than that, we measure the inlet, outlet pressure. We measure the flow rate, the voltage and current. So we've fully instrumented the setup to get a highly accurate view of the performance of this device. Once we tested it, we first evaluated at this lower flow rate. The, so we're able to reach 1.4 kilowatts. The typical industry standard is 1.5 liters per minute per kilowatt, so that equates to flow rate at 2.2 liters per minute. And this is the test data from that flow rate, with a, horizontal orientation and let temperature it wandered a little bit as the experiment progressed, but we started at 35 Celsius, and it was 38 Celsius right at that highest heat load. So we're able to push it up to 175 watts per square centimeter in 100 watt steps. The OHP, you can kind of see the the red line is the, is that thermocouple nine that I referenced on that previous slide. So you can see some slight oscillations or temperature oscillations. That's due to the fluid motion in the oscillating heat pipe. It's inherent in the oscillating heat pipe. It varies by design. And with this design, it was, one to two degrees Celsius temperature oscillation. Our temperature difference at that 1,400 watts was 28.4 degrees Celsius. Okay. So we, converting all of our test data into resistances, we tested at a range of flow rates. You can see the lowest was 2.2 liters per minute. We tested at five and six point one liters per minute. Although, 6.1 was a single data point. We only tested at the highest power. And we also tested at the 5.1 liter per minute, flow rate. We tested 23 Celsius, 35 Celsius, 53 Celsius inlet temperatures, so test range inlet temperatures. You could see on that first chart the flow rate variation that barely affects the performance of the Aussling heat pipe. It's pretty much negligible. The resistance is identical no matter what flow rate you're pushing through. For the inlet temperature, there is a slight performance effect. But once you get above about 500, maybe 700 watts, the difference is minimal. But at the lower operating temperature, you do get slightly higher performance, and that's partially due to the shift in properties for each EW and partially due to the OHP's inherent performance. In addition, it's not shown on the slide, but we tested it in a side and horizontal orientation and performance was identical. So there's no orientation effect. And we were, going into this, we didn't expect to see a difference and we found no difference. The so backing off and just looking at the flow side of the, oscillating heat pipe. And here you kind of have a cross sectional view. You can see that the fin array area is this course grid. Again, it's millimeter gap between the fins. And we're testing with EGW, which has a relatively high viscosity. It's 86% higher than the p g 25. With that, we were able to, well, at 2.2 liters per minute, our pressure drop was less than point two PSI, so very low pressure drop. And as we increase the flow rate, we never exceeded you can see this highest pressure drop at, was 5.1 liters per minute or sorry, that was the seven liters per minute did not exceed 1.4, psi. So the OHP across the board had a very low pressure drop and this is because of a very coarse thin structure, but as I'll show in a second, the performance is identical with other high end cold plates. So we have a thermal force that's identical with a much lower pressure drop, and that's due to that spreading in the oscillating heat pipe, enabling a much larger fin array. So we also let's say, comparing the oscillating heat pipe to the same identical cold plate without the actual oscillating heat pipe structure in it. So we manufactured a second one, and it just had solid aluminum where the awesome heat pipe channels, would exist. And that, the OHP's performance was about two x or 2.8 times the conductance of that, cold plate without the oscillating heat pipe and it tested at 5.5 liters per minute with inlet temperature of 35 Celsius. So, you know, as as a baseline check, we were verifying that the OHP was performing better. Comparing the Austin heat pipe to the limits of operation, we overlay the experimental data against the limits of operation, which I described earlier. It tells us where the OHP will and will not work. So, as you can see the vapor inertia limit, which is applicable in this case, seplink limit is not. Vapor inertia limit is the one that we constrain the operation of the Aspen heat pipe. However, with our current test test, we were not close to that limit. Our test bed, had the 1,400 watt and 80 Celsius, upper limits. However, across the board, the heat pipe worked perfectly, without any issues whatsoever. So we're well away from the limit and the performance was excellent throughout. We then compared our experimental results versus some published academic results. The citations are on the last or second to last slide. So the Ausling heat pipe or our work are the red triangles in the in both graphs. As you can see on the upper graph, we're comparing our normalized flow rate that's liters per minute per kilowatt to the thermal resistance. So that's the, the two axes. And the, you know, the the Ausling heat pipe OHP, cold plate was as good as the other published works that we're comparing against. However, or addition, the pressure drop was substantially lower than those in the published works. So with our current first generation FIND design, we met the high performance of published results and drastically exceeded the, pressure drop of the published results. We can, iterate and refine the fin design to optimize for higher, heat rejection into the working fluid so that thermal resistance will drop with a redesign with a tighter fin structure. It will also increase the pressure drop. However, we have a substantial margin on the pressure drop from existing cold plates. So depending on the application, there are several knobs we can turn to optimize for the, to make the perfect oscillating pipe cold plate for that application. In conclusion, the oscillating pipe cold plate, we developed it in aluminum, sized for a typical cold plate, nine point or, you know, data center cold plate. It's 9.8 centimeters by 6.8 centimeters, 2.6 centimeters thick. Our mass was two hundred and seventy five grams, which is substantially lower than the typical copper cold plate out there in the industry. The performance at that, sort of data center standard, 2.2 liters per minute when you're at 1.4 kilowatts. The performance was point zero two kelvins per watt or point one six four kelvins in the square per watt. This was with ethylene glycol water which has lower performance than propylene glycol 25. So when tested with that other working fluid, we would expect the calcium heat pipe performance to increase, by a decent margin. And this OHP cold plate matched the performance of existing high performance cold plates, and the pressure drop is over 30 times lower than existing high performance cold plates. The OHP cold plates was lighter, a 3.3 times lighter than the equivalent copper cold plate. Alright. Any questions? Hey, Corey. Thanks for sharing about, what your team at ThermoVont has been developing. We are ready to dive into some questions. And it isn't too late to ask your questions. So, as I said earlier, type them into the q and a tab to the right of the chat, and we will do our best to get to all of them. Corey, ready to go? Sure. Absolutely. I'm gonna put these on the screen so you'll be able to see them as well. Let's share the first one. It says, can this device be adapted to other chip sizes? Yes. So this was a initial demonstration, oscillating heat pipe cold plate. We sized it for a relatively small chip, but there are no reasons why this could not be adapted too much to a larger chip or a different, heat input pattern. Great. Alright. Moving on. The next question asks, is there a reason why copper wasn't used for this device? So there are okay. The Ausling heat pipe, we have a working fluid you put in there. I did not describe it, but, there are some compatibility issues with, copper. We can change the working fluid in the oscillate heat pipe, which would enable a copper device. So it would require a little bit of retooling of design, but it is possible for us to change that. Performance might be a little bit different. We'd have to evaluate that. Great. The next question says, what is the dry out limit for this configuration? Does the active liquid cooling help achieve higher limits for this? And what is the cooling channel structure of the cold plate? Or is the increase in area of dissipation computed solely based on the heat pipe and design? I'm gonna leave that up for a minute so you can absorb all of that. Okay. Yes. A few different questions. Dry out limit, let's say I I present that on the slide there. Let's say, 80 Celsius, we should be able to reach, it was around, oh, 2,600, watts. It was approximately that number. I had it written on one of the previous slides. And then, we should be able to also push up at the same heat load at 1,400 watts. You should be able to push up to, the mid 80 Celsius range. So that was the the limit, but you could see there's a curve. So I just picked two points on that curve. If you feel free to reach out, I can talk to, you in more detail. And then does the active liquid cooling help achieve higher limits for this? The housing heat pipes performance or its limits are let's see, when we model those limits, they are taking to account the active cooling. So, yes, it has shifted higher because of that, but those the limits as drawn are for the, for the cooling as we have tested. And was the cooling channel structure. I showed that in some slides. It was that one millimeter gap with half millimeter, fin. The pattern is sort of a grid pattern. It's slightly staggered in the middle, but it's a relatively simplistic structure. Great. And was And and. I and I do wanna say, we you shared your, presentation with us. If if you go to the docs tab, you can download a PDF of Corey's slides. Thank you. Let's get to the next one. Next question. Ask, what is the expected result if you used Skype to FINS? If we use Skyfins, the, thermal performance should increase substantially. The arsenic heat pipe will spread to a much to that large rejection area into the sky fins, and the, the pressure drop will rise substantially also. Now since we still have a much larger ejection area, so we have a much larger cross sectional flow, our pressure drop probably is not as high as a typical sky fin cold plate, but it will be higher than what you're seeing. And the thermal performance will increase, I would expect by 20 ish percent, but that's just a rough estimate. Great. Next question says, is this current OHP automotive application ready? We haven't done, specific, you know, certification for automotive application, but the typical, requirements for automotive are similar to our typical, aerospace, shock and vibe and, you know, it's it's not quite identical. There's a little bit different there, but we would expect that the Austin heat pipe would be automotive ready, you know, with the requirements of, you know, in the automotive industry. Great. I've got it looks like the next two questions are, materials related. The next one asks why aluminum sixty sixty one was used. So it was partially due to the fluid we decided to test in there. We also primarily manufacturing aluminum sixty sixty one. It's not the only material we manufacture in, but with the aerospace industry where weight is a concern, that is our, typical material. But, we can, you know, evaluate this design and other materials as well. Sure. You know, maybe go for some titanium or something. Yeah. Exactly. Wish we all had titanium budgets. Right? Next question asks, could other patterns other than fins be interesting? Think freeform in technical ceramics. Yes. They could. We also we do have to integrate it with the oscillating heat pipe that wraps around it. So? whatever manufacturing technique would need to either survive our current braze aluminum braising process or, we'd need a different process. So there'd be some manufacturing development that'd be required, depending on how complex that geometry is. Great. I see a comment from Siddharth. Siddharth says, not a question. Convey my thanks to Corey for his detailed answers. So, you're doing a great job. Looks like we have, one question left. And if you still have more questions, type them into the q and a, and we'll make sure that, Corey and his team get them after the event, and he'll follow-up. Last question I have here in the chat, Corey, is this. It says, at which level do you have, 175, watts per centimeter and eighty eighty degrees Celsius operating temperature at junction or case? I I don't know if I got all that exactly right. Yeah. So we were measuring at case temperature with that thermocouple nine, that was we used a thermal grease. The thermocouple nine was, with had a little bit of, silicone backing, so it's pressed onto the case of the Asling heat pipe. You know, there might have been a tiny amount of grease between the interface, but it should be, accurately measuring the case temperature and not the junction. So the junction would be higher than that. Great. Corey, any closing thoughts? We were chatting before, and it sounds like ThermoVont's doing some really cool research and development right now. So you've got a a really cool job, and, obviously, you know your stuff. What do what do you wanna leave us with today? We're we're excited to, you know, see where this syncing pipe technology, it's primarily what we develop. We're excited to see what applications this can be used in because there's a, we think a wide range of unique applications for the Auslan heat pipe that we can really, would really benefit with an Auslan heat pipe. Awesome. Well, make sure to take advantage of, if you wanna download Corey's slides, they're in the docs tab. I just dropped the website for Thermavont in the chat. It's thermavont.com. And, Corey, thanks for being with us today at data center cooling days. Thank you very much. Awesome. That's gonna wrap it up for this session. Our next one will begin at 11:30AM eastern, which is in about fifteen minutes, and we're gonna hear from Allison Bowen at Open Compute Project. So we will see you then.